31_Michel_J.F._Digonnet_Rapid_Prototyping_of_Digita_428_2008

234 Rapid Prototyping of Digital Systems Chapter 12

Mark (1) ASCII “P” = 0x50
Space (0) 00 001010

Start 0 1 23 45 6 7 Stop
Data Bit number
Bit Bit
LSB Time MSB

Figure 12.2 RS-232C Serial interface transmission of an 8-bit data value

Figure 12.2 shows the transmission of a single ASCII character over an RS-
232C serial interface. The serial bit has two states. Mark is the high state (>3V)
and Space is the low state (<-3V). Older generation serial devices will have
around +12V and -12V levels for Mark and Space. Note that for the proper RS-
232 voltage levels, a standard digital logic output bit will have to have its
voltage levels converted for use in a serial interface. Special ICs are normally
used for this RS-232C voltage conversion. To reduce the need for additional
circuits, these ICs also generate the required DC supply voltages from the
standard digital logic DC power supplies. This special IC chip is already
present on the UP3’s serial interface. FPGA logic elements can be used to build
the UART hardware function.

The idle state is High (Mark). Whenever the interface starts sending a new 8-bit
data value, the line is dropped Low (Space) for one clock cycle (baud rate
clock). This is called the start bit. The eight data bits are then clocked out
during the next eight baud clocks in low to high bit order. The highest data bit
is sometimes used as a parity bit for error detection, when only seven data bits
are used instead of eight. After the data bits are clocked out, the bit goes high
for one clock. This is called the Stop bit. Sometimes at low baud rates, two
Stop bits are present. Note that at least 10 clocks are required to transfer an 8-
bit data value.

Typically, UARTs transfer 8-bit data values in and out to other internal logic
using an 8-bit parallel I/O port interfaced to a processor. Extra UART status
bits can be read by the processor that indicate another 8-bit data value can be
sent to the UART or another 8-bit data value is available to read in from the
UART. Since serial transmission is very slow compared to a processor’s clock,
these status bits must be checked in software or hardware for their proper state
or the processor will send/receive data faster than the UART can produce or
consume it. Other status bits can also be used to detect various error conditions.

Legacy Digital I/O Interface Standards 235

A UARTs transmitter uses a shift register clocked by the baud rate clock to
convert the 8-bit parallel data to eight serial bits. Start and stop bits are
automatically added by the UART’s hardware.

At the other end of the serial cable, another UART’s receiver uses the Stop and
Start bits to reset its internal state machine that is attempting to synchronize its
receive clock phase to the incoming serial bit data. This state machine
synchronizes the receive clock phase whenever it sees an edge on the incoming
serial line. Note that several consecutive bits could be the same value inside the
eight data bits, so there is not an edge transition on every single clock.

A UART typically uses an internal clock that is eight or sixteen times the baud
rate to watch for edges on the incoming serial data line. UARTs also use this
faster clock and a counter to attempt to sample the data bits in the middle of
each bit’s time frame to minimize the possibility of reading in an incorrect
value near an edge. Since long wires are allowed on an RS-232C serial
interface, there will likely be noise and ringing present whenever the serial bit
changes. Clocking in the bit in the middle of its time frame greatly increases
the reliability of the interface. A second shift register is used for serial to
parallel conversion in the UART’s receiver circuit.

Some serial devices also require additional hardware handshake lines to stop
and start the flow of a new 8-bit data value over the serial interface. These
handshake lines require additional signal wires in the cable used to connect the
serial device. Some of the more commonly used handshake lines are RTS
(request to send), CTS (clear to send), DCD (data carrier detect), DSR (data set
ready), and DTR (data terminal ready).

There are two types of serial devices defined in the RS-232C standard, data
terminal equipment (DTE) and data carrier equipment (DCE). A standard RS-
232C serial cable is designed to connect a DTE device to a DCE device. When
connecting two serial devices of the same type, a special null modem cable or
adapter is needed. A null modem exchanges the TD and RD signal lines at one
end of the cable along with several connections on the handshake lines. If you
experience problems when connecting a new serial device, the various
handshake and null modem cable options can be quickly checked using a low-
cost in-line RS-232C analyzer breakout box.

The UP3 board contains a RS-232 serial connector, and it has the required
voltage conversion IC needed for serial data transmission.

12.3 SPI Bus Interface

The serial peripheral interface (SPI) bus created by Motorola in the 1980s is
used primarily for synchronous serial communication between a host processor
and peripheral ICs. Four signal lines are used: Chip Select (CS), Serial Data
Input (SDI), Serial Data Output (SDO), Serial Clock (SCLK). CS and SCLK
are outputs provided by the master device. The slave devices receive their clock
and chip select inputs from the master. If an SPI device is not selected, its SDO
output line goes into a high impedance state (tri-state). The number of serial
bits transferred to the slave device varies from device to device. Each slave
device contains an internal shift register used to transfer data.

236 Rapid Prototyping of Digital Systems Chapter 12

Two types on connections between master and slave devices are supported as
seen in Figure 12.3. In a cascaded connection, all slaves in the chain share a
single chip select line driven by the master. The master device outputs SDO and
it connects as an input to a slave device’s SDI input. A slave’s SDO output
connects to another slave’s SDI input. The serial data cascades through all of
the slaves and the final slave in the chain connects its SDO line to the master’s
SDI input to complete the chain. In this configuration, the slave devices appear
as one larger slave device, the data output of one device feeds into the input of
another device, thus forming one large shift register.

Master Slave Master Slave 0
SDI SDO SDI SDO
CS0 CS0
CS1 SDI CS1 Slave 1
Slave CS2 SDI SDO
CS2
SDO Slave 2
SCLK SCLK SDI SDO
SDO SDI SDO
Slave
SDI SDI
SDO

Figure 12.3 The two SPI slave device configuration options.

The second SPI configuration option supports independent slave devices, each
device has its own unique chip select input line coming from the master. The
master’s SDO output connects to each slaves SDI input. The slave’s SDO tri-
state outputs are connected together and to the master’s SDI input. Only the
selected slave’s SDO output is driven, the others are tri-stated.

Multiple masters are also supported in SPI. Several SPI modes are supported
with serial data being valid on either the rising edge or the falling edge of the
clock. Serial clock rates can range from 30 kHz to 3 MHz depending on the
devices used. Most commonly, devices place new data on the bus during the
falling clock edge and data is latched off the bus on the rising edge after it
stabilizes, but you will need to check data sheets for specific master and slave
devices to confirm this since some devices use the opposite clock edges.

Some Motorola literature may use different names for the SPI signals. CS may
appear as SS, SDI as MOSI, and SDO as MISO. In National Semiconductor
products, SPI is also known as Microwire. SPI devices are also available in
several different voltage supply levels ranging from 2.3 to 5 volts. Since SPI
uses a common clock, the hardware interface is simpler than RS-232C serial.

Legacy Digital I/O Interface Standards 237

12.4 I2C Bus Interface
The Inter IC (I2C) bus is a widely used standard developed by Phillips in the
1980s for connecting ICs on the same circuit board. Many small ICs now
include I2C pins to transfer data serially to other ICs. For lower bandwidth
signals, a serial interface has an advantage in that it requires fewer interconnect
lines. The I2C bus uses two signal wires called SCL and SDA. SCL is the clock
line and SDA is the 1-bit serial data & address line. A common ground signal is
also needed. The SCL and SDA lines are open drain. This means that the output
is only driven Low, never High. An external pull-up resistor pulls the lines
High whenever there is not a device driving the lines Low.

In an FPGA with tri-state output pins, you can simulate open drain outputs by
tri-stating the output whenever the bit should go High and only driving the
output signal Low. Even though there are multiple devices on the I2C bus, only
one pull-up resistor is used for the entire I2C bus.
Devices on the I2C bus are masters or slaves. The slaves are the devices that
respond to bus requests from the master. Each slave is assigned its own unique
7-bit I2C bus address. Since both address and data information is transferred
over the bus, the protocol is a bit more involved than SPI. When the master
needs to talk to a slave, it issues a start sequence on the I2C bus. In a start
sequence, SDA goes from High to Low while SCL is High. To stop an I2C
sequence, the master sends a stop sequence command. In a stop sequence, SDA
goes from Low to High while SCL is High. Start and stop sequences are the
only times a change may occur in SDA while SCL is High.
The master drives the SCL clock line to transfer each new I2C serial bit. To
force a wait, a slave device can drive SCL Low. Therefore, before each new I2C
SCL clock, the master checks to see if SCL is being forced Low by a slave. If it
is, the master must wait. SCL clocks are typically up to 100 kHz with 400 kHz
available on some new devices.

SD A MSB ACK Signal ACK Signal from
from Receiver Receiver
SCL 12 Byte Complete
S Addr es s Clock Line held
Low while serviced
START
7 89 1 2 3-7 8 9
R/W ACK
ACK P

Data STOP

Figure 12.4 I2C interface serial transmission of an 8-bit data value

All address and data transfers contain eight bits with a final acknowledge
(ACK) handshake bit for a total of nine bits. All address and data transfers
send the High bits first, one per SCL clock bit High. In an address transfer, the

238 Rapid Prototyping of Digital Systems Chapter 12

7-bit address is sent and the eighth bit is a R/W bit (0=read, 1=write). Some IC
datasheets just append this final R/W bit to the address field and show an 8-bit
address field (with even 8-bit addresses for read and odd for write).
The last bit in all data and address transfers, bit nine, is an ACK from the slave.
The slave normally drives ACK Low on the last SCL cycle to indicate it is
ready for another byte. If ACK is not Low, the master should send a stop
sequence to terminate the transfer.
As seen in Figure 12.4, when a master wants to write data to a slave device, it
issues the following bus transactions:

1. Master sends a start sequence.
2. Master sends the 7-bit I2C address (high bits first) of the slave with the

R/W bit set Low.
3. Master sends the 8-bit internal register number to write.
4. Master sends 8-bit data value(s). Highest bits first.
5. Master sends a stop sequence.

When a master wants to read data from a slave device, it issues the following
bus transactions:

1. Master sends a start sequence.
2. Master sends the 7-bit I2C address of the slave (high bits first) with the

R/W bit set Low.
3. Master sends the 8-bit internal register number to read.
4. Master sends a start sequence.
5. Master sends the 7-bit I2C address of the slave (high bits first) with the

R/W bit set High.
6. Master reads the 8-bit data value(s). Highest bits first.
7. Master sends a stop sequence.

In the full I2C standard, multiple bus masters are also supported with collision
detection and bus arbitration. Collision occurs when two masters attempt to
drive the bus at the same time. Arbitration schemes must decide which device
can drive the bus when multiple masters are present. Some of the newest I2C
devices can support a high-speed 3.4 MHz clock rate, 10-bit addresses,
programmable slave addresses, and lower supply voltages. The System
Management Bus (SMB) bus developed by Intel in 1995 that is used for
temperature, fan speed, and voltage measurements on many PC motherboards is
based on the I2C bus. On the UP3 board, the real-time clock chip and the serial
EEPROM chip use an I2C bus interface. Many new TVs, automobiles, and other
consumer electronics also contain I2C interfaces between chips for control
features.

Legacy Digital I/O Interface Standards 239

SPI and I2C both offer good support for communication with low-speed
devices. SPI is better suited to applications that need to transfer higher
bandwidth data streams without the need for explicit address information. Some
of the most common SPI examples are analog-to-digital (A/D) and digital-to-
analog (D/A) converters used to continuously sample or output analog signals.
Since addressing is required for I2C, it requires more hardware, but with
advances in VLSI technology these additional hardware costs are minimal. In
2005, one FPGA vendor calculated that a single I/O pin on an FPGA package
costs as much as 50,000 transistors inside the chip.

12.5 For Additional Information
The books Parallel Port Complete and Serial Port Complete by Jan Axelson
published by Lakeview Research (www.lvr.com) contain complete details on
using parallel and serial ports. The full I2C specification is available from
Philips Semiconductors (www.phillipssemiconductor.com) and SMB at
(www.smbus.org). The Motorola MC68HC11 data manual
(www.freescale.com) and various National Semiconductor manuals
(www.national.com) have more information on SPI. Analog Devices
(www.analogdevices.com) makes a wide variety of A/D and D/A converters
with SPI and parallel interfaces.

12.6 Laboratory Exercises

1. Interface a printer with a parallel port to the UP3 board’s parallel port. Connect the two
devices using a printer cable. Design logic using a state machine or a processor core for
the FPGA to transfer data and handle the handshake lines. You may want to use an older
printer so that any problems with your design will not damage the printer. Be careful not
to generate tri-state bus conflicts on the parallel data lines by making sure you drive the
data direction bit to the proper state. Have the UP3 print a short ASCII message on the
printer ending with an ASCII form or page feed to print the message on a page. A form
feed may be needed to cause the printer to print since most printers store characters in an
internal page buffer.

2. Interface the FPGA board’s serial port (DE1, DE2 or UP3) to a PC serial port using a
serial cable. Run a serial communications program on the PC. Send a short message to
the PC from the FPGA and display the data from the PC on the FPGA board’s LCD
panel or seven-segment LEDs.

3. Design an I2C interface for the UP3 board’s real-time clock chip. Display the time from
the chip on the UP3 board’s LCD display. Don’t forget to check the UP3 board’s jumper
settings and battery for the real-time clock chip. The data sheet for the clock chip
contains address and data formats.

240 Rapid Prototyping of Digital Systems Chapter 12

4. Obtain an IC chip with an SPI or I2C interface and design an interface for it on the FPGA
board. Chips with SPI interfaces include analog-to-digital converters, digital-to-analog
converters and various sensor modules. Header I/O and power connections are available
on the FPGA boards with 5V or 3V logic levels. Consult the board’s user manual for pin
assignments.

CHAPTER 13
FPGA Robotics
Projects

Photo: The FPGA-bot is a small robot controlled by an FPGA board

242 Rapid Prototyping of Digital Systems Chapter 13

13 FPGA Robotics Projects

13.1 The FPGA-bot Design

The FPGA-bot shown in Figure 13.1 is a low-cost moving robotics platform
designed for use the DE1, DE2, UP3, or UP2 board. The FPGA-bot is designed
to be a small autonomous vehicle that is programmed to move in response to
sensory input. A wide variety of sensors can be easily attached to the FPGA-
bot.

The round platform is cut from plastic and a readily available 7.2V or 8.4V R/C
rechargeable battery pack is used to supply power. Two diametrically opposed
drive motors move the robot. A third inactive castor wheel or skid is used to
provide stability. The robot can move forward, reverse, and rotate in place. Two
relatively inexpensive radio control servos are used as drive motors. The FPGA
is programmed to act as the controller. The R/C servos are modified to act as
drive motors. The servos are controlled by timing pulses produced by the FPGA
board.

Figure 13.1 The FPGA-bot uses an R/C car battery and R/C servos for drive motors.

13.2 FPGA-bot Servo Drive Motors
A typical radio control servo is shown in Figure 13.2. Servos have a drive
wheel that is controlled by a coded signal. The servo shown is a Futaba S3003
which is identical, internally, to the Tower TS53J servo. Radio control
servomotors are mass-produced for the hobby market and are therefore
relatively inexpensive and consistently available. They are ideally suited for
robotics applications. Internally, the servo contains a DC drive motor (seen on

FPGA Robotics Projects 243

the left in Figure 13.2), built-in control circuitry, and a gear reduction system.
They are small, produce a relatively large amount of torque for their size, and
run at the appropriate speed for a robotics drive motor.

Figure 13.2 Left: Radio Control Servo Motor and Right: Servo with Case and Gears Removed.

The control circuitry of the servo uses a potentiometer (variable resistor) that is
used to sense the angular position of the output shaft. The potentiometer is the
tall component on the right in Figure 13.2. The output shaft of a servo normally
travels 180-210 degrees. A single control bit is used to specify the angular
position of the shaft. The timing of this bit specifies the angular position for the
shaft. The potentiometer senses the angle, and if the shaft is not at the correct
angle, the internal control circuit turns the motor in the correct direction until
the desired angle is sensed.

The control signal bit specifies the desired angle. The desired angle is encoded
using pulse width modulation (PWM). The width of the active high pulse varies
from 1-2 ms. A 1ms pulse is 0 degrees, 1.5ms is 90 degrees and a 2 ms pulse is
approximately 180 degrees. New timing pulses are sent to the servo every 20
ms.

13.3 Modifying the Servos to make Drive Motors

Normally, a servo has a mechanical stop that prevents it from traveling move
than half a revolution. If this stop is removed along with other modifications to
the potentiometer, a servo can be converted to a continuously rotating drive
motor. Modifications to the servo are not reversible and they will void the
warranty. Some robot kit vendors sell servos that are already modified.

To modify the servo, open the housing by removing the screws and carefully
note the location of the gears, so that they can be reassembled later. The
potentiometer can be replaced with two 2.2K ohm ¼ watt resistors or
disconnected by cutting the potentiometer shaft shorter and setting it to the

244 Rapid Prototyping of Digital Systems Chapter 13

center position so that it reports the 90-degree position. A more accurate setting
can be achieved by sending the servo a 1.5ms pulse and adjusting the
potentiometer until the motor stops moving. The potentiometer can then be
glued in place with CA glue. In the center position the potentiometer will have
the same resistance from each of the outside pins to the center pin. If the
potentiometer is replaced with two resistors, a resistor is connected between
each of the two outside pins and the center pin.

In some servos, there will be less mechanical play if the potentiometer is
disabled by cutting the center pin and modified by drilling out the stop on the
potentiometer so that it can rotate freely. The two resistors are then added to
replace the potentiometer in the circuit.

The largest gear in the gear train that drives the output shaft normally has a tab
molded on it that serves as the mechanical stop. After removing the screw on
the output shaft and removing the large gear, the mechanical stop can be
carefully trimmed off with a hobby saw, knife, or small rotary-grinding tool.
The servo is then carefully re-assembled.

After modifications, if a pulse shorter than 1.5 ms is sent, the motor will
continuously rotate in one direction. If a pulse longer than 1.5 ms is sent the
motor will continuously rotate in the other direction. The 1.5 ms or 90-degree
position is sometimes called the neutral position or dead zone. The drive signal
to the motor is proportional, so the farther it is from the neutral position the
faster it moves. This can be used to control the speed of the motor if the neutral
position is carefully adjusted. A pulse width of 0 ms or no pulse will stop the
servomotor.

A servo has three wires, +4 to +6 Volt DC power, ground, and the signal wire.
The assignment of the three signals on the connector varies among different
servo manufacturers. For Futaba servos, the red wire is +5, black is ground, and
the white or yellow wire is the pulse width signal line. For JR and Hitec servos,
the orange or yellow wire is the signal line and red is +5, and black or brown is
ground.

On the FPGA-bot, the FPGA board must be programmed to provide the two
timing signals to control the servo drive motors.

13.4 VHDL Servo Driver Code for the FPGA-bot

To drive the motors a servo signal must be sent every 20 ms with a 0, 1, or 2 ms
pulse. The FPGA board is programmed to produce the timing signals that drive
the motors. If no pulse is sent, the motor stops. If a 1 ms pulse is sent, the
motor moves clockwise and if a 2 ms pulse is sent the motor moves in the
reverse direction, counterclockwise. To move the FPGA-bot forward, one motor
moves clockwise while the other motor moves counterclockwise. This is
because of the way the motors are mounted to the FPGA-bot base.

In the code that follows, lmotor_dir and rmotor_dir specify the direction for the
left and right motor. If both signals are ‘1’ the UP3 bot moves forward. The
VHDL code actually moves one motor in the opposite direction to move
forward. If both are ‘0’ the robot moves in reverse. If one is ‘1’ and the other is
‘0’, the UP3 bot turns by rotating in place. The two speed controls are

FPGA Robotics Projects 245

lmotor_speed and rmotor_speed. In the speed control signals, ‘0’ is stop and ‘1’
is run. A 1kHz clock is used for the counters in the module. The FPGAcore
function, clk_div, can be used to provide this signal. Two more complex
techniques for implementing variable speed control are discussed in problems
at the end of the chapter. Acroname sells a low-cost optical encoder kit made by
Nubotics that can be attached to standard R/C servo wheels and used for
position feedback and more accurate motor speed control.

LIBRARY IEEE; : IN STD_LOGIC;
USE IEEE.STD_LOGIC_1164.ALL; : IN STD_LOGIC;
USE IEEE.STD_LOGIC_ARITH.ALL; : IN STD_LOGIC;
USE IEEE.STD_LOGIC_UNSIGNED.ALL; : OUT STD_LOGIC);
ENTITY motor_control IS

PORT (clock_1kHz
lmotor_dir, rmotor_dir
lmotor_speed, rmotor_speed
lmotor, rmotor

END motor_control;

ARCHITECTURE a OF motor_control IS
SIGNAL count_motor: STD_LOGIC_VECTOR( 4 DOWNTO 0 );

BEGIN
PROCESS
BEGIN
-- Count_motor is a 20ms timer
WAIT UNTIL clock_1kHz'EVENT AND clock_1kHz = '1';
IF count_motor /=19 THEN
count_motor <= count_motor + 1;
ELSE
count_motor <= "00000";
END IF;
IF count_motor >= 17 AND count_motor < 18 THEN
-- Don’t generate any pulse for speed = 0
IF lmotor_speed = '0' THEN
lmotor <= '0';
ELSE
lmotor <= '1';
END IF;
IF rmotor_speed = '0' THEN
rmotor <= '0';
ELSE
rmotor <= '1';
END IF;
-- Generate a 1 or 2ms pulse for each motor
-- depending on direction
-- reverse directions between the two motors because
-- of servo mounting on the FPGA-bot base
ELSIF count_motor >=18 AND count_motor <19 THEN
IF lmotor_speed /= '0' THEN
CASE lmotor_dir IS
-- FORWARD

246 Rapid Prototyping of Digital Systems Chapter 13

ELSE WHEN '0' =>
END IF; lmotor <= '1';

-- REVERSE
WHEN '1' =>

lmotor <= '0';
WHEN OTHERS => NULL;
END CASE;
ELSE
lmotor <= '0';
END IF;
IF rmotor_speed /= '0' THEN
CASE rmotor_dir IS
-- FORWARD
WHEN '1' =>

rmotor <= '1';
-- REVERSE
WHEN '0' =>

rmotor <= '0';
WHEN OTHERS => NULL;
END CASE;
ELSE
rmotor <= '0';
END IF;

lmotor <= '0';
rmotor <= '0';

END PROCESS;
END a;

13.5 Low-cost Sensors for an FPGA Robot Project

A wide variety a sensors can be attached to the FPGA board. A few of the more
interesting sensors are described here. These include infrared modules to avoid
objects, track lines, and support communication between FPGA-bots. Other
modules include sonar and IR to measure the distance to the nearest object and
a digital compass to determine the orientation of the FPGA-bot. Most robots
will need to combine or “fuse” data from several types of sensors to provide
more reliable operation.

Signal conditioning circuits are required in many cases to convert the signals to
digital logic levels for interfacing to the digital inputs and outputs on the FPGA
board. Analog sensors will require an analog-to-digital converter IC to interface
to the FPGA board, so these devices pose a more challenging problem. Small
low-cost A/D ICs are available with SPI interfaces that require a minimal
number of FPGA pins.

Sensor module kits are available and are the easiest to use since they come with
a small printed circuit board to connect the parts. Sensors can also be built
using component parts and assembled on a small protoboard attached to the
FPGA-bot. Sensor modules are interfaced by connecting jumper wires to digital
inputs and outputs on the FPGA board’s J3 and J2 expansion header connector.

FPGA Robotics Projects 247

Sensors with a single output bit can utilize a simple control scheme, and for
basic tasks the robot can be controlled using hardware as simple as a state
machine. More advanced sensors that report actual distance, location, or
heading measurements will likely require a processor core on the FPGA
running a program that interprets sensor readings and implements the robot’s
control algorithm.

• Line Tracker Sensor

A line tracker module from Lynxmotion is shown in Figure 13.3. This device
uses three pairs of red LEDs and infrared (IR) phototransistor sensors that
indicate the presence or absence of a black line below each sensor. When the
correct voltages are applied in a circuit, an IR phototransistor operates as a
switch. When IR is present the switch turns on and when no IR is present the
switch turns off. The LED transmits red light that contains enough IR to trigger
the phototransistor.

Each LED and phototransistor in a pair are mounted so that the light from the
LED bounces off the floor and back to the IR phototransistor. The LED and IR
sensor must be mounted very close to the floor for reliable operation. Black
tape or a black marker is used to draw a line on the floor. The black line does
not reflect light so no IR signal is returned. Three pairs of LEDs and IR
phototransistor sensors produce the three digital signals, left, center, and right.
The FPGA-bot can be programmed to follow a line on the floor by using these
three signals to steer the robot. The mail delivery robots used in large office
buildings use a similar technique to follow lines or signal cables in the floor.

Figure 13.3 – Three LEDs and phototransistors are mounted on bottom of the Line Tracker board.

• Infrared Proximity Detector
An IR proximity sensor module from Lynxmotion is seen in Figure 13.4. The
FPGA-bot can be outfitted with an infrared proximity detector that is activated
by two off-angle infrared transmitting LEDs. The circuit utilizes a center-

248 Rapid Prototyping of Digital Systems Chapter 13

placed infrared sensor (Sharp GP1U5) to detect the infrared LED return as seen
in Figure 13.5. The Sharp GP1U5 was originally designed to be used as the IR
receiver in TV and VCR remote control units. From the diagram, one can see
that the sensitivity of the sensor is based on the angle of the LEDs. The LEDs
can be outfitted with short heat-shrink tubes to better direct the infrared light
forward. This prevents a significant number of false reflections coming from
the floor. The IR sensor will still occasionally detect a few false returns and it
will function more reliably with some hardware or software filtering.

Figure 13.4 IR Proximity Sensor Module – Two IR LEDs on sides and one IR sensor in middle.

Figure 13.5 Proximity detector active sensor area.

FPGA Robotics Projects 249

As seen in Figure 13.6, the circuit on the IR proximity module utilizes a small
feedback oscillator to set up a transmit frequency that can be easily detected by
the detector module. This module utilizes a band-pass filter that essentially
filters out ambient light. Some older first generation electronic ballasts used in
commercial fluorescent lights can interfere with the IR sensors since they
operate at the same frequency as the filter. Newer ballasts now operate at a
higher frequency since they also caused problems with IR TV remote control
signals.

38 KHz 180
LED Enable.H Ohm
Signal Detect.L
1 of 2
IR LEDs
5v

IR Detector

Figure 13.6 Circuit layout of one LED and the receiver module on the infrared detector.

In Figure 13.6, when the Left_LED Enable signal is High, the Low side of the
IR LED is pulled to ground. This forces a voltage drop across the LED at the
frequency of the 5v to ground oscillating signal. In other words, the LED
produces IR light pulses at 38 kHz. Using a 38 kHz signal helps reduce noise
from other ambient light sources.

Since the IR detector has an internal band-pass filter centered at 38 kHz, the
detector is most sensitive to the transmitted oscillating light. The 5v pull-up
resistor allows the IR Detector’s open collector output to pull up the SOUT
signal to High when no IR output is sensed. To detect right and left differences,
the right and left LEDs are alternately switched so that the detected signals are
not ambiguous. If both the left and right LEDs detect an object at the same
time, the object is in front of the sensor.

If the IR sensor was built from component parts, a hardware timer implemented
on the UP3 board could be used to supply the 38-40 kHz signal. Similar IR
LEDs and IR detector modules are available from Radio Shack, #276-137B,
and Digikey, #160-1060. Assuming two FPGA-bots are equipped with IR
sensor modules, it is also possible to use this module as a serial communication
link between the robots. One FPGA-bot transmits using its IR LED and the
other FPGA-bot receives it using its IR sensor. To prevent interference, the IR
LEDs are turned off on the FPGA-bot acting as a receiver. Just like an IR TV

250 Rapid Prototyping of Digital Systems Chapter 13

remote, the IR LED and sensor must be facing each other. Bandwidth is limited
by the 38kHz modulation on the IR signal and the filters inside the IR detector.
(An IR sensor strip that converts IR to visible light is available from Radio
Shack. This sensor can be used to confirm the operation of IR LEDs.)

• Wheel Encoder

The Nubotics WW01 WheelWatcher incremental quadrature encoder system
from Acroname is shown in Figure 13.7. This low-cost electronics board bolts
onto the top of a standard-size R/C servo. The adhesive-backed codewheel
attaches to a wheel mounted on the servo’s output shaft. Two pairs of optical
emitters and receivers bounce light beams off of the codewheel.

Note that there are 32 black stripes on the reflective codewheel. When the
wheel is rotating, the encoder produces two series of digital pulses that are 90
degrees out of phase. When one of the pulses changes twice before the other
pulse changes, the direction has been reversed. 128 clock pulses per revolution
are produced and a separate direction signal indicates the current direction of
rotation. By counting pulses with a counter or by accurately measuring the time
between individual pulses using a fast hardware counter on the FPGA, it is
possible to more accurately control the position and velocity of the servo motor.
When used on robot drive motors, this optical encoder feedback provides more
accurate position and speed control for the robot.

Figure 13.7 Nubotics WW-01 Wheel Watcher Incremental Encoder System.

FPGA Robotics Projects 251

• Sonar Ranging Units

The Devantech SRF10 Sonar Module is shown in Figure 13.8. This device uses
ultrasonic sound waves to measure distances from a few inches to around 35
feet. They are widely used in robotics. The timing of the sound echo indicates
distance to the nearest object. The transducer first functions as a transmitter by
emitting several cycles of a ultrasonic signal, and then functions as a receiver to
detect sound waves returned by bouncing off nearby objects. Even though
ultrasound is inaudible, the transducer also generates a slight audible click each
time the device transmits. The beamwidth is rather wide, and several sonar
modules facing in different directions are commonly used.

The time it takes for the ultrasonic echo signal to return is measured using an
IC mounted on the back side of the board. This time is converted to distance
since sound travels out and back at 0.9 ms per foot. Only around 10-20 samples
per second are possible with the device since it takes time to wait for echoes to
return. Some sonar modules require external hardware to measure the pulse
timing to produce the distance to target. The SRF10 device operates off +5V
DC, and it sends distance measurements back to the host using an I2C bus. A
number of other similar Sonar modules are available.

Figure 13.8 Devantech SRF10 Ultrasonic Range Finder.

• IR Distance Sensors
The Sharp GPD2D02 seen in Figure 13.9 is an IR device that can provide
distance measurements similar to the slightly more expensive sonar sensor. This
sensor has a shorter range of 10 to 80 cm (~ 4 to 32 inches). The distance is
output by the sensor on a single pin as a digital 8-bit serial stream.

252 Rapid Prototyping of Digital Systems Chapter 13

Figure 13.9 Sharp IR Ranging Module. Far Object

Scattered
Reflection

Lens Near Object

IR LED Lens

Position
Sensitive
IR Detector

Figure 13.10 Operation of Sharp IR Ranging Module.

As shown in Figure 13.10, internally the GPD2D02 contains an IR LED and a
position-sensitive IR detector. The IR LED transmits a modulated beam of
infrared light. When the light strikes an object, most of the light will be
reflected back to the LED. Since no surface is a perfect optical reflector,
scattering of the IR beam occurs at the surface of the object and some of the
light is reflected back to the position sensitive detector. By comparing the near
and far object beams shown in Figure 13.10, it is apparent that the position at
which the scattered reflected IR beam hits the detector is a function of the
reflection angle.

FPGA Robotics Projects 253

The 8-bit integer value reported by the sensor in cm is approximately

1000 ∗ tan−1⎜⎛ 1.9 ⎞⎟+ offset .
⎝ DISTANCE ⎠

The constant 1.9 is the distance between the lenses in cm. The offset is the no-
object present value returned by the sensor. This offset constant can vary by as
much as 17 between different sensors and has a typical value of 25. Note that a
close object reports a larger value and a distant object reports a smaller value.
Objects closer than 10cm will report an incorrect value and should be avoided
by placing the sensor away from the edge of the robot. Large objects beyond 80
cm can sometimes report an incorrect value that makes them appear closer.

A special connector (Japan Solderless Terminal #S4B-ZR) is required to
connect to the GPD2D02. If desoldering equipment is available, the small
connector can also be desoldered from the sensor and wires attached directly to
the sensor.

In addition to +5V and ground pins, the sensor has an input, Vin, and a serial
output, Vout. Vin is an input to the sensor that clocks out the serial data on
Vout. When Vin is Low for around 70 ms, the sensor takes a reading. When a
reading is available, Vout goes High.

On each of the next eight falling clock edges of Vin, the sensor will output a
new data bit. The eight data bits should be clocked into the FPGA on the rising
edges of Vin (when they are stable). When clocking out the data, the clock
period on Vin should be 0.4 ms or less. The eight data bits are clocked out in
high to low order. If Vin is not dropped Low within 1.5 ms after clocking out
the final data bit, the sensor shuts down to save power. A shift register can be
used to assemble the data bits. The demo program ir_dist.bdf on the DVD
contains a VHDL-based IP core for use with the GP2D02 sensor.

The sensor’s Vin pin is an open-drain input. Open-drain or open-collector
inputs should never be driven High. An FPGA’s tri-state output pin can be
connected directly to an open-drain input, if the tri-state output is never driven
High. When Vin should be High, tri-state the FPGA’s output pin and when the
output should be Low, drive the output pin Low with the tri-state gate turned on
with a low output.

Open-drain or open-collector inputs contain an internal pull-up resistor to +5V.
Multiple open-drain (open-collector) outputs can be tied together to a single
open drain (open-collector) input to perform a wired-AND operation. Any one
of the outputs can pull the input Low. If no output pulls the signal Low, a single
pull-up resistor forces the input High. This wired-AND operation occurs just by
tying the open-drain (open-collector) outputs together and no physical AND
gate is needed. In negative logic, a wired-OR operation occurs.

Normal gate outputs cannot be connected. This wired-AND logic only works
because these gates have special output circuits that do not contain a transistor
that forces the input High. This transistor is present in normal gate outputs. If a
normal gate output is connected to other open-drain (open-collector) outputs,

254 Rapid Prototyping of Digital Systems Chapter 13

its transistor could turn on to force the input High at the same time another
gate’s output transistor turns on to force it Low. This would short the power
supply to ground drawing excessive current that might damage the devices. An
analog and longer range version of this IR distance sensor are also available.

• Magnetic Compass Sensors

Various electronic components are available that detect the magnetic field of
the earth to indicate direction. A low-cost digital compass sensor is shown in
Figure 13.11. The Dinsmore model 1490, often used in electronic automobile
compasses, is a combination of a miniature rotor jewel suspended with four
Hall-effect (magnetic) switches. Four active-low outputs are provided for the
four compass directions. When the module is facing North, the North output is
Low and the other three outputs will be High. Eight directions are detected by
the device, since two outputs can become active simultaneously. In this way,
the device can indicate the four intermediate directions, NE, SE, SW, and NW.
NE for example activates the active-low North and East outputs. The device
can operate off +5V.

Mount any compass device as far away from motors as possible to avoid
magnetic interference from the magnets inside the motor. Four 2.2K ohm pull-
up resistors to +5V are required to interface to the UP3 board, since the four
digital output pins, N, S, E, and W, all have open-collector outputs. Just like a
real compass, a time delay is needed after a quick rotation to allow the outputs
to stabilize. If the compass module leads are carefully bent, the compass
module and the four required pull-up resistors can be mounted on a standard
20-pin DIP, machined-pin, wire-wrap socket and connected to the UP3 header
socket.

An analog version of the device is available with 1-degree accuracy, but it
requires an analog-to-digital conversion chip or signal phase timing for
interfacing.

Figure 13.11 Dinsmore 1490 Digital Compass Sensor.

FPGA Robotics Projects 255

• Electronic Compass Sensors

Low-cost electronic compass modules are also available that detect the
magnetic field of the earth to indicate direction. The cost is two to three times
that of the mechanical compass described in the previous section. New
generation electronic compass modules offer more accuracy and faster settling
times than mechanical compass sensors. An electronic compass module from
PNI is shown in Fig. 13.12. This module contains a 2-axis magneto-inductive
sensor and an ASIC. Heading information and magnetic field measurement data
is available using a digital SPI serial interface.

Figure 13.12 PNI Electronic Compass Module.

• Low-cost Gyros and Accelerometers

Gyros and accelerometers are useful sensors for robots that need a balance
sense. This can include robots that balance on two wheels like the Segway
Human Transporter, robots that walk on two legs, and even robots that fly.
Gyros and accelerometers have traditionally been used in aircraft autopilots and
inertial measurement units (IMUs). Helicopters use a gyro to stabilize and
control the tail rotor. Recently, Microelectromechanical Systems (MEMS)
technology has produced small low-cost piezo-gyroscope and accelerometer
ICs. These devices were originally used in automobile airbags. The gyros
output a voltage level that is proportional to the speed or rate of the tilt angle
changes. An analog-to-digital converter will be needed to input the gyro signal.
The MEMS accelerometers output a voltage level or a pulse that changes its

256 Rapid Prototyping of Digital Systems Chapter 13

duty cycle proportionally (e.g., PWM) to the tilt angle by sensing the change in
acceleration due to gravity. Gyros will drift slowly over time and an
accelerometer is needed to correct the gyro’s drift. Without an accelerometer to
correct for gyro drift, the tilt error slowly grows to the point where the robot
would lose its balance. Accelerometers will respond more slowly to tilt than the
gyro, so both a gyro and accelerometer is typically needed for each axis that
needs a balance sense.

A complementary filter is used to combine or fuse sensor data from both the
gyro and accelerometer to generate a more accurate tilt angle. Kalman filtering
techniques can be used to improve the accuracy of noisy measurements. Noise
levels are still somewhat high at very low G forces on these low-cost gyros and
accelerometer IC sensors, so currently they are not useful for navigation since
they cannot accurately determine the exact location of a slow moving robot by
integrating the sensor measurements over time.

Analog Devices makes a variety of these sensors and sells small evaluation
boards for them. It is likely that small low-cost sensor modules containing both
a MEMS gyro and an accelerometer with a microcontroller will be available
commercially in the near term.

Figure 13.13 Small sensor board for an aircraft autopilot system. Photograph ©2004 courtesy of
Henrik Christophersen , Georgia Institute of Technology Unmanned Aerial Research Facility.

FPGA Robotics Projects 257

Figure 13.13 shows a sensor board for an autopilot system that is used for
unmanned aircraft. In the top corner, three MEMS gyros and accelerometers are
mounted at right angles to provide data on all three axes. The white square flat
module between the two vertical assemblies is a GPS receiver. The black
square ICs at each end of the vertical assemblies are airspeed and altitude
sensors. An A/D chip with an SPI interface is used to read sensors that have
analog voltage outputs. The three square modules near the bottom edge of the
board are DC to DC voltage converters. The lower board contains an FPGA and
a DSP processor.

• GPS and DGPS receivers

The Global Positioning System was built by the US Department of Defense to
provide highly precise worldwide positioning. Triangulation using radio signals
from several satellites provides a position accurate to 25 meters. With an
additional land-based correction signal, Differential GPS (DGPS) improves the
accuracy to 3 meters. DGPS receivers provide ideal position data for robot
navigation. Unfortunately, with current systems you are not likely to receive the
GPS radio signals indoors in most buildings, so their use is typically limited to
larger more rugged outdoor robots. Low-cost single chip GPS modules such as
the Motorola FS Oncore seen in Figure 13.14 or the Ublox in Figure 13.13 are
currently available. An SPI serial interface is supported. A new generation of
highly sensitive GPS systems is being developed that may function indoors in
some buildings.

Figure 13.14 Motorola Single Chip GPS module.

• Thermal Image Sensors

Low-cost thermal image sensors can provide thermal imaging data for robots.
Most thermal sensors such as those used in motion detectors and burglar alarms
detect only movement. Thermopile sensors measure the temperature of a heat
source. One such sensor, the Devantech TPA81 Thermopile Array is shown in
Figure 13.15. It contains 8 Pyro-electric sensors arranged in a column that

258 Rapid Prototyping of Digital Systems Chapter 13

detect infra-red in the radiant heat range of 2um to 22um range. It contains an
on-board PIC microcontroller.

When the sensor is mounted on a servo, it can be used to horizontally scan an
area and generate a thermal image. Candle flames and human body heat can be
detected several feet away at room temperature. It uses an I2C bus for
interfacing to the host controller.

Figure 13.15 Devantech TPA81 Eight Pixel Thermal Array Sensor.

• Solid State Cameras
Low-cost solid state cameras can provide visual sensors for robots. Keep in
mind that advanced image processing and visual pattern recognition requires
complex algorithms that need a lot of processing power. The CMUCAM2
developed at Carnegie Mellon University seen in Figure 13.16 contains a PIC
microcontroller and can transfer image data using a serial connection. It can
track color blobs and report their location and size in an image at 26 to 50
frames per second.
Low-cost USB cameras are another option, but they will require a USB core
interface and additional image processing. The low-cost CMOS color camera
assembly OV6620 or OV7620 used in the CMUCAM2 module from
Omnivision (www.ovt.com) can also be directly interfaced to an FPGA. It uses
an I2C interface for camera control signals and a separate parallel bus is used to
transfer image data.

FPGA Robotics Projects 259

Figure 13.16 The CMUCAM2 contains a color video camera on a chip and a microcontroller.

13.6 Assembly of the FPGA-bot Body

Assembly of the FPGA-bot can be accomplished in about an hour. A drill or
drill press, screwdriver, scissors, a soldering iron, and a wire stripper are the
only tools required. First, obtain the parts in the parts list. Next, drill out the
holes in the round Plexiglas base (part #15) as shown in Figure 13.17. The
mounting holes will move around depending on which FPGA board is used.
The DE2 board is a bit larger than the other boards and you may want to
increase the size of the plastic base to accommodate it or mount it on long
standoffs well above the servo wheels.
To prevent scratches, leave the paper covering on the Plexiglas until all of the
holes are marked and drilled out. The front of the base is on the right side in
Figure 13.17. The wheel slots are symmetric with respect to the center of the
circle.

Proper alignment of the four screw mounting holes for the FPGA board is
critical. Unscrew the four standoffs from the bottom of the FPGA board.
Carefully place it towards the rear of the plastic base as shown in Figure 13.17,
and mark the location of the screw holes using a pen or pencil. Any FPGA
board can also be used, but the mounting holes will be in different locations.
Leave extra space in front of the FPGA board on the plastic base for use by
forward facing sensor modules as seen in Figures 13.1 and 13.9. Double check
that the board clears the top of the servo wheels once it is mounted. Longer
standoffs can also be used to clear the wheels, if needed. Locate the cable and
switch holes as shown in Figure 13.17. Exact positioning on these holes is not
critical. If one is available, use an automatic center punch to help align the drill
holes. The board’s expansion header pins should face towards the front of the

260 Rapid Prototyping of Digital Systems Chapter 13

base so that it is easy to attach the sensors. Re-attach the standoffs to the FPGA
board and set it aside. After all holes are drilled, remove the paper covering the
Plexiglas.

1 3/8”

Switch and Wheel Slots Servo Cable
Power Cable 3/8” by 2 3/4” Two 3/8”
Two 7/32” Holes
Holes

UP3 Mounting Screw
Four 1/8” Holes

1 3/8”

3/16” Plexiglass 10.5” Circle

Figure 13.17 FPGA-bot Plexiglas Base with wheel slots and approximate drill hole locations for the
UP3 board. Verify the exact dimensions and mounting hole locations for each individual FPGA
board.

Mount the toggle switch (part #8) in the hole provided in the base. If available,
Loctite or CA glue can be used on the switch mounting threads to prevent the
switch nut from working lose. Solder the red wire (+7.2V) from the battery
connector to one of the switch contacts. This is the connector with wires that
plugs into the battery pack connector (part #4). Solder one of the twin lead
wires (part #9) to the other switch terminal. Solder the other twin lead wire to
the black (GND) battery connector wire and insulate the splice with heat shrink
tubing or electrical tape (part #10).
Route the twin lead wire through the hole provided in the base. A small knot in
the twin lead on the bottom side of the base can be used for strain relief. Solder
the power connector (part #11) to the other end of the twin lead wire on the top
of the base. The center conductor is +7.2V and the outer conductor is ground on
the power connector.

FPGA Robotics Projects 261

Check the power connections with an ohmmeter for shorts and proper polarity
before connecting the battery. For strain relief and extra insulation, consider
sealing up the power connector with Silicone RTV or insulating one of the wire
connections with heat shrink tubing. Be careful, NiCAD and NiMH batteries
have been known to explode or catch on fire, if there is a short. A fuse on the
battery power wire might be a good idea, if you are prone to shorting out
circuits.

Figure 13.18 Bottom view of FPGA-bot base showing battery, servos, wheels, and cabling.

Attach the battery pack (part #2) to the bottom of the Plexiglas base with
sticky-back Velcro (part #16). Figure 13.18 is a close-up photo of the bottom
side of the FPGA-bot. A NiMH battery pack is shown in Figure 13.18. If you
use a larger NiCAD battery pack, it can be mounted in the middle of the base
about one inch off center towards the rear wheel, with the battery pack
connector facing the rear.

262 Rapid Prototyping of Digital Systems Chapter 13

The battery is moved towards the rear for balance to place the weight on the
rear skid. The Velcro on the base should be around 2 inches longer than the
battery pack towards the rear of the robot to allow for positioning of the battery
later on to balance the robot. The wire and connector on the battery pack should
also be attached to the base to prevent it from dragging on the floor. Attach a
small piece of Velcro on the rear of the connector so that the battery wires can
be attached to the base. Attach the battery pack to the base.

On UP3, UP2, and UP1 boards, solder a 60-pin female header socket (part #7)
to the expansion header B location. Attach the FPGA board to the top of the
base with 4-40 screws (part #18), using the hex spacers provided on the FPGA
board (part #14). Figure 13.19 is a close-up photo of the top of the FPGA-bot.
Double check power connections and polarity with an ohmmeter. The inner
contact on the power connector should be +7.2V, the outer contact is ground,
and the toggle switch should turn it off. Then plug the power connector into the
FPGA board. Plug in the battery connector and flip the power switch. An LED
should light up on the FPGA board indicating power on. The Cyclone
expansion B header socket faces the front of the robot. DE2 and DE1 boards
already have expansion sockets soldered to the board.

Mount the wheels (part #5) on two modified servos (part #3). If you are not
using the special servo wheels, you may need to enlarge the hole in the center
of each wheel by drilling it out partially with a drill bit that is the same size as
the servo output shaft. The depth of the hole should be slightly shorter than the
servo output shaft and not all the way through the wheel, so that the wheel does
not contact the servo body. The servo output shaft screw is inserted on the side
of the wheel with the smaller hole. A washer may be required on the servo
screw. The wheel should not contact the servo case and must be mounted so
that it is straight on the servo. CA glue or Blue Loctite can also be used to
attach the wheels and screws more securely to the servo output shaft.

Attach the servos to the bottom of the base using double sided foam tape (part
#17) or a more durable servo mounting bracket. The servo body faces toward
the center of the base. Be sure to carefully center the wheels in the plastic-base
wheel slot. If you are using foam tape, make sure all surfaces are clean and free
of grease, so that the foam tape adhesive will work properly. Lightly sanding
the servo case and adding a drop of CA glue helps with tape adhesion. Route
the servo connector and wire through the holes provided in the base.

Attach a tail wheel to the base or a skid (part #19) at the rear of the battery
pack using layers of foam tape as needed. Move the battery as needed so that
the robot has proper balance and rests on the two wheels and the rear skid.

Attach another skid to the front of the battery pack using several layers of foam
tape. The front skid should not contact the floor and at least ¼ inch of clearance
is recommended. The front skid only serves to prevent the robot from tipping
forward during abrupt stops.

On the UP3, attach a 3-pin .1 inch header (part #9) to the small wire wrap
protoboard in an open area. One is required for each servo on the robot. Solder
wires from the appropriate pins J2 and J3 connections on the protoboard to the
new header pins. The three wires on the servo are Vcc (4.8 to 6 volts), ground,

FPGA Robotics Projects 263

and the PCM control signal wire. Some manufacturers’ servos have different
power connections, but they all have three pins. Wrap extra servo wire around
the hex spacers underneath the UP3 board.

Figure 13.19 Top View of FPGA-bot Base with Compass, IR, and Sonar Sensor Modules. A UP3
FPGA board is shown, but any FPGA board can be used.

Optional sensor modules such as the IR proximity detector or line tracker can
be attached to the base unit with foam tape. Run wires from the sensors to the
expansion connectors. A small .1 inch wire wrap protoboard with 40-pin female
header connectors soldered to the protoboard as shown in Figure 13.19 is handy
for making servo and sensor connections to the FPGA board. In Figure 13.19, a
third servo is used to make a sensor turret for IR and Sonar distance sensors.

• Parts List for the FPGA-bot

1. An Altera FPGA Board. The FPGA board serves as the controller for the FPGA-bot. It
is attached to the FPGA-bot body with screws. No modifications are required to the
board. Any of the Altera FPGA boards can be used, but mounting holes will be in
different locations.

264 Rapid Prototyping of Digital Systems Chapter 13

• Parts Available from a Hobby Store

2. A 7.2V to 8.4V 1300-1700mAh Rechargeable NiCAD battery pack with the
standard Kyosho battery connector. This is a standard R/C car part, and it is used to
power the FPGA-bot. For a small additional cost, new NiMH batteries are also available
that store almost twice the energy per weight. The battery will need to be charged prior
to first use.

3. Two modified R/C Servomotors. Two identical model servos are required so that the
motors run at the same speed. Servo modifications are described in section 13.3. Any
servo should work. The following servos have been tested: Tower Hobbies TS53J,
Futaba S148 and S3003, and HS 300. Some manufacturers’ servos appear to run in the
reverse direction. This is easily fixed in the hardware design since the motor controller
is implemented on the UP3 board. Several robot parts vendor sell modified servos for a
slightly higher cost. Ball bearing servos are worth the extra cost, if you intend to run the
robot constantly for several months. A third unmodified servo will be needed if you
want a rotating sensor turret as shown on the example FPGA-bot photo.

4. Kyosho Female Battery connector with wire leads, Duratrax or Tower Hobbies
#DTXC2280. This is used to connect to battery. A connector is needed so that the
battery can be disconnected from the FPGA-bot and connected to a charger.

5. Two Acroname or Lynxmotion servo wheels. These wheels are 2 ¾ plastic wheels
that are designed to attach to the servo’s output shaft spline. Prather Products 2¼-inch
aluminum racing wheels with rubber O ring tires, Tower Hobbies #PRAQ1810 or Hayes
Products #114, 2 ¼-inch hard plastic racing wheels (also available from Tower Hobbies)
can be used as a substitute. These somewhat smaller two alternative wheels will work,
but they do not have the spline to match the servo output shaft and are a bit more
difficult to connect reliably than the Acroname or Lynxmotion servo wheels.

6. A Castering Wheel or Two small Teflon or Nylon Furniture Slides. There is a bit too
much mechanical play in common furniture casters for a small robot and they tend to
randomly deflect the robots direction after sharp turns. Lynxmotion’s #TWA-01 is a
mini castering robot tail wheel built using an R/C airplane tail wheel that works well.
The mounting wire needs to be bent a little off center so that the wheel quickly rotates to
the direction of travel. Other robot parts vendors such as Acroname also have robot tail
wheels, but a spacer may be required to adjust the height. The battery will need to be
moved a bit and perhaps rotated ninety degrees to accommodate them and still maintain
proper balance on the robot base. Magic Sliders 7/8-inch diameter circular discs also
work well on flat surfaces. The slides are used as a skid instead of a third wheel on the
FPGA-bot. Metal or hard plastic will also work. Attached to the bottom of the battery
with several layers of foam tape, an optional front skid can be used for stability during
abrupt stops. On flat surfaces, a Teflon skid actually works better than a common small
furniture caster from a hardware store.

FPGA Robotics Projects 265

7. A charger for the 7.2V or 8.4V battery pack. An adjustable DC power supply can be
used to charge the battery if it is properly adjusted and timed so that the battery is not
overcharged. Overcharged batteries will get hot and will have a shorter life. Automatic
peak-detection quick chargers are the easiest and most foolproof to use. These chargers
shut off automatically when the battery is charged. One quick charger can be used for
several robots as a full charge is achieved in less than 30 minutes with around 5 Amps
maximum charge current. Inexpensive trickle battery chargers deliver only around 75
mA of charge current, and they will require several hours charge the battery.

• Parts Available from an Electronics Parts Store
8. Three 40-pin .1-inch double row PC board mount female header sockets, DigiKey
#S4310 or equivalent. These sockets are soldered into a small 0.1” center wire wrap
protoboard that fits into the Santa Cruz Expansion connector on the UP3. This is used to
connect servos and sensors to the UP3 board.

9. A 2 to 3 inch strip of .1” single row breakaway headers. DigiKey #S1021-36 or
equivalent These headers are used to make custom servo and sensor connectors on the
protoboard. They can be soldered to the protoboard.

10. A small wire wrap protoboard with holes on .1” centers cut down to 2” by 2.8”. A
This is used to make a protoboard for use with the UP3 board. The protoboard contains
connectors for servos and sensor. A protoboard with solder pads makes it easier to
mount the connectors.

11. A miniature toggle switch with solder lug connections. The switch should have a
contact rating of more than two amps (Radio Shack #275-635B or equivalent). Only two
contacts or single pole single throw (SPST) is needed on the switch to turn power on
and off. If all of your servos and sensors connect to the UP3 and do not use the
Vunregulated supply, you could eliminate the switch by using the UP3’s power switch.

12. Approximately 9 inches of small-gauge twin-lead speaker wire. This part is used to
connect power to the UP3 board. The wire must fit into the DC power plug (part# 11).
Typically, 20-22 gauge wire is required. Two individual wires can also be used, but twin
lead is preferred.

13. A 1-inch piece of small heat shrink tubing or electrical tape. This part is used to
insulate a splice in the twin-lead power wire.

14. A Coaxial DC Power Plug with 5mm O.D. and 2.1mm I.D., Radio Shack Number
274-1567 or equivalent. This power plug fits the power socket on the UP3 board. A
different size plug is needed for the UP2 board, use #274-1568 that has a 2.5mm I.D.

15. An assortment of small wire jumpers and connectors to attach wires to the male
headers on the UP3. These are the jumper wires commonly used for protoboards. Two

266 Rapid Prototyping of Digital Systems Chapter 13

short jumpers are used to connect the two servo signal wires, and other jumpers are used
to any connect sensor boards.

16. Four, 1-inch hex spacers with 4-40 threads or use the shorter spacers that come
with the board. These are used to mount the UP3 board to the Plexiglas base using the
holes in the UP3 board.

• Parts Available from a Hardware Store

17. 3/16-inch thick Plexiglas cut into a 10.5-inch diameter circle. This part is the base of
the robot. Colored Plexiglas such as opaque white, will not show scratches as easy as
clear. Holes to cutout and drill are shown in Figure 13.11. If a band saw, jig saw, or
other machine tool is not available, a local plastics fabricator can cut this out. When
using a number of very large sensors or a DE2 board, it may be necessary to increase the
size slightly or add another circular deck for sensor mounting. A larger robot requires
more space for maneuvering. To prevent scratches on the Plexiglas, keep the paper
backing on the plastic until all of the holes have been marked and drilled out. The size
of the wheel slots may need to change depending on the wheels you select.

18. One 8-inch long strip of 2-inch wide sticky-back Velcro. Two 8-inch long strips, 1
inch wide can also be used. The Velcro is used to attach the battery to the bottom of the
Plexiglas base. Since the battery is attached with Velcro and a connector, it can be
quickly replaced and removed for charging.

19. Approximately 8 inches of 1-inch wide double-sided 3M foam tape. This is used to
attach servos, skids, and optional sensor boards to the base. Be sure to clean surfaces to
remove any grease or oil prior to application of the tape for better adhesion. For a more
durable servo mount, Lynxmotion has aluminum servo mounting brackets that can be
used instead of the double sided tape.

20. Four 4-40 Screws 5/16-inch or slightly longer. The screws are used to attach the UP3
board to Plexiglas. The screws thread into the hex spacers attached to the UP3 board.

21. Blue Loctite, Cyanoacrylate (CA) Glue, and Clear Silicone RTV. These adhesives
and glues are useful to secure screws, servos, and wheels. The mechanical vibration on
moving robots tends to shake parts loose over time. These items can also be found at
most hobby shops. Only a few drops are needed for a single robot. A single tube or
container will build several robots.

13.7 I/O Connections to the board’s Expansion Headers

Most servos and sensor I/O signals will need to be attached to an expansion
header. Many FPGA pins are now 3.3V. R/C Servos and most sensors use 5V,
but be sure to check the device’s datasheet. Don’t forget to connect a ground
signal between the device and the FPGA board, even if the device has it’s own
power supply or a direct connection to a battery. Several ground and power pins

FPGA Robotics Projects 267

are available on the FPGA board’s expansion headers. A 5V 1A power supply
pin is available on the UP3’s J2 expansion connector. J4 has a 3.3V supply
connection pin and JP6 can be used as another 5V supply connection.

A small protoboard can be built to connect servos and sensors to the FPGA
board. All of the connectors and pins on the boards line up on tenth inch
centers. A 0.1” perfboard or wire wrap protoboard can be cut down to 2” by 2
7/8” so that it fits over J1, J2, J3, and J4. 0.1” 40 pin connectors to attached the
protoboard to connect to J1..4 can be mounted on the protoboard. A wire warp
protoboard with holes every .1” has solder pads that can be used to attach
connectors using solder. Point to point wiring and soldering can be used to
make connections on the protoboard from the J1..J4 connectors to the .1”
connectors used to attach servos and sensors. Small single row strips of .1”
header pins can be snapped apart to make male connectors on the board for the
servos and most sensors.

You may want to consider isolating your robot’s servo or motor power supply
from the supply used for the FPGA board’s logic to control the noise generated
on the supply lines by the DC motors. On larger robots, two batteries are
sometimes used. A Vunregulated connection that does not go through the 5V
regulator and is connected directly to the UP3’s power input jack is available
on JP8 and J4. The 9V supply is connected after the input power switch on the
UP3 and to JP5. This also can be used to power servos and motors, assuming
the battery voltage level is not too high. If the battery voltage is too high,
another regulator can be used for the motors.

At a minimum, decoupling capacitors connected across the servo’s power
supply connections are a good idea. If you plan on having several sensors on
your robot, you may want to consider building a small PCB with header pins
for the sensor power and data connections as seen in Figure 13.19. Most R/C
servos can run on 4.8 to 6V.

13.8 Robot Projects Based on R/C Toys, Models, and Robot Kits

A second option for building an FPGA driven robot involves modifying a low-
cost radio-controlled (R/C) car or truck. Fundamentally, almost any large R/C
car or truck can be modified to work with the Altera board, although some are
clearly better choices than others.

In our robot, we used a Radio Shack (www.radioshack.com) R/C 4WD SUV
shown in Figure 13.20. The R/C platform affords a more robust drive train and
control; however, turning radius and noise levels are sacrificed over the smaller
FPGA-bot. The R/C SUV has a spring suspension and large soft tires that make
it operable outdoors on rougher surfaces. Following are some R/C car selection
considerations that will affect available modifications and control of the new
platform.

268 Rapid Prototyping of Digital Systems Chapter 13

Figure 13.20 FPGA Controlled Toy R/C Truck with IR Distance Sensors.

• Seven-Function Controls

When choosing an R/C car, select one that has a remote control with at least
seven remote functions (forward, backward, forward-right, forward-left,
backward-right, backward-left, and stop). Note that these low-cost R/C cars do
not have variable speed or variable turning controls; however, once they are
interfaced to the FPGA board, variable speed and turning can be accomplished
by changing the duty cycle of the command signals. (More on this later.)

A control module built using the FPGAs logic allows a relatively inexpensive
R/C car to perform with the capabilities of the more expensive cars with
“digital proportional steering” and “digital proportional speed controls.” Once
interfaced to the FPGA board, an IP core (Robot_CTL) is used to handle
control of all direction and speed control outputs. As illustrated in Figure 13.21,
the IP core control module affords a higher degree of control than the original
radio control. The outputs connect to the R/C cars internal control circuits that
drive the DC Motors.

FPGA Robotics Projects 269

FwdRev 1 Bit 0 = Forward/1 = Reverse
Direction 3 Bits First bit Left/Right, 2nd and 3rd bit is angle.
0-00 = Left – Straight*
Speed 3 Bits 0-01 = Left – Slight Turn
0-10 = Left – Medium Turn
0-11 = Left – Full Turn
1-00 = Right – Straight*
1-01 = Right – Slight Turn
1-10 = Right – Medium Turn
1-11 = Right – Full Turn
* Note: 000 and 100 are both Straight
000 = Stop
001 = Slowest Speed

:::

111 = Fastest Speed

Figure 13.21 Robot Control IP Core with Pulsed Speed & Steering Control.

• Speed

When considering the speed of the vehicle, a modest speed is more desirable
than the faster speeds. At 800 feet per minute, our prototype FPGA controlled
robot car moves fast enough to be difficult to catch. In almost all cases, the
robot is operated at half the maximum speed or less. The limiting factor is
generally the delay inherent in the sensor’s input sampling rate and range. A
fast moving car typically will hit the wall before a collision sensor can take the
data samples needed to initiate avoidance.

To control the speed (or the degree of turn) a repeated pulse train is sent to the
forward or reverse signals (left or right signals for direction). Instead of a
steady high signal causing the car to move forward at full speed, the pulse train
varies the duty cycle to change speed (or degree of turn). By modulating the
duty cycle of the pulse train, “digital proportional control” can be implemented
on each control signal. In other words, changing the duty cycle can control the
speed and the degree of turn. The more the duty cycle approaches 100%, the
harder the turn and/or the faster the speed.

270 Rapid Prototyping of Digital Systems Chapter 13

The frequency of the pulsed control signal used must be higher than the natural
mechanical frequency response of the system. A very slow changing pulse will
cause the motor and gears to vibrate and make additional noise. Pulse
frequencies of a few kHz are typically used to avoid this problem.

When reversing direction on a moving DC motor, it is common practice to
include a small time delay with the motor turned off to reduce the inductive
voltage spikes produced by the motor windings. Recall that changing current
flow through an inductor produces voltage. Without the delay in some circuits,
these high voltage spikes can damage or reduce the life of the transistors
controlling the motor. This delay can be incorporated in the IP control core, if
needed.

Figure 13.22 illustrates the relationship between turn angle, speed, and duty
cycle on the control signals. The figure implies that the duty cycle is linear, i.e.,
a 50% duty cycle produces half speed. Actually, the duty cycle is very non-
linear and highly dependent on the type of car, size of the DC motors, and
power. (An R/C car with a dying battery performs as if the duty cycle is
considerably less.) By experimenting with patterns of the 16-bit speed and
direction vectors used in the IP core controller, a more linear relationship can
be established between the command bit patterns and the actual performance of
the vehicle.

Turn Speed
Angle

Full

1/2

1/3

1/4

Figure 13.22 Affect of Duty Cycle on Turning Angle and Speed.

• Battery Choice

The choice of car will also dictate the type of batteries and charger that will be
needed. Note that some cars come with 9.6V packs and others come with 7.2V
packs. Both should work well with the UP3 board as a controller. The prototype
used the 7.2V pack that discharged quickly and required a second pack placed
in parallel with the first to support longer run times. The cars with a 9.6V pack
should give the UP3 board’s 5V onboard regulator a better regulator margin and
a longer life between recharges.

FPGA Robotics Projects 271

• Mounting the FPGA Board

Before you select an R/C car, make sure that there is a good place to mount the
FPGA board. If the car is large enough, there is usually a large flat area under
the car body cover molding to secure the FPGA board.

• Interfacing the FPGA Board to the R/C Car

Remove appropriate body cover screws and expose the PC board receiver and
control module. Most current low-cost R/C toy cars have a single electronic PC
board that contains both control circuits. Generally, there is one 16 or 18-pin
DIP radio command demodulator chip in the center of the board that converts
the radio signals into simple digital control signals. These digital control
signals then activate the H-bridge circuit that controls the DC motors that drive
the wheels of the R/C car.

An H-bridge is a standard electronic circuit used to control DC motors. It
allows for both forward and reverse operation of the same DC motor. H-bridge
circuits contain four large power transistors that are needed to turn on and
reverse a DC motor. Discrete transistors may be used to build the H-bridge or it
may be in an IC or packaged module that connects directly to the motors.

NEVER ATTEMPT TO DRIVE A MOTOR OR RELAY WITH AN FPGA PIN DIRECTLY, IT CANNOT
SUPPLY THE HIGH CURRENT LEVELS THAT THESE DEVICES NEED. THE FPGA’S OUTPUT PIN

DRIVER CIRCUIT MAY BE DESTROYED.

If the car supports seven functions, it will have at least four pins coming off of
the DIP chip package that break down into Left, Right, Forward, and Reverse.
Using a voltmeter or an oscilloscope, test which pins change when the remote
control is set to each of the four directions. From each of the designated
command pins on the chip, the trace on the PC board will run to separate H-
bridge circuits for each motor.

By clipping or desoldering and pulling out the four control pins on the chip
going to the board and soldering wires from each chip pin hole pad trace to the
FPGA (Figures 13.23 and 13.24), the FPGA board can control the four
directions and speed of the R/C car using the car’s existing H-bridge circuits. In
our modification, we desoldered the entire chip and put a socket on the board.
To have the original control signals coming from the radio control module also
sent to the FPGA board, run four more wires from the clipped chip pins to the
one of the FPGA’s headers. If the four clipped chip pins (RCx) are connected to
the FPGA board, logic can be designed to include the original radio control
functions supplied by the handheld remote control unit. You can also make this
connection at the H-bridge circuit if necessary. On the UP3, you will want to
use the UP3s 5V I/O pins for this interface.

272 Rapid Prototyping of Digital Systems Chapter 13

RCLt
RCRt

Rt RCRev
RCFwd
Lt
Rev
Fwd

Figure 13.23 Interfacing to the R/C Car’s Internal Control Signals at the Demodulator IC.

Figure 13.24 Photo Showing Control Modifications to R/C Car Control Board.

• Hobbyist R/C Models, Robot Kits, and Commercial Robot Bases

For those with a larger budget, higher quality R/C hobbyist cars are available
with built-in proportional steering and pulsed electronic speed controls. The
control interface to these cars uses standard R/C PWM signals that are identical
to the PWM servo control bit described at the end of Section 13.2. An example
R/C Hummer can be seen in Figure 13.25. This robot is controlled using a C
program running on the FPGA’s Nios processor core. Various robot kits without
control electronics or a computer are also available. Almost any of these robot
kits can be controlled by the UP2 or UP3 board provided they are large enough
to carry it and can power it from their battery or carry a second battery for the

FPGA Robotics Projects 273

UP3. An interesting walking robot kit containing 12 R/C servos is seen in
Figure 13.26.

Figure 13.25 Hobbyist R/C model with a CMU camera and R/C PWM servos controlled by an FPGA
Figure 13.26 Lynxmotion Hexpod Walking Robot Kit with 12 R/C servos

274 Rapid Prototyping of Digital Systems Chapter 13

Figure 13.27 ActiveMedia’s Amigobot robot base controlled by an FPGA with a Nios Processor

Some commercial robot bases are also available such as the Amigobot as seen
in Figure 13.27, the ER1 from Evolution Robotics and the mobile robot
platform from Drrobot. The Amigobot uses an RS-232C serial interface and the
ER1 uses USB for motor control. A robot base contains a motor, drive
electronics, sensors, and a battery, but it has no high-level controller.

One of the newest and most attractive low-cost options for a robot base is
iRobot’s iCreate robot base as seen in Figure 13.28 (www.irobot.com). It is
basically a “Roomba” robotic vacuum cleaner without the vacuum parts. It
contains an internal microcontroller, two drive motors with feedback, audio
output, and several simple sensors (i.e., IR, bump or contact switches, and IR
cliff or surface drop-off sensors). Using an RS-232 serial port, motor
commands can be sent to the microcontroller and the sensor status can be read
back. An FPGA board running a Nios processor can talk to the iCreate
microcontroller using the RS-232 serial port. You can write the required robot
application code for the Nios processor in C.

6-32 screw holes are provided on top of the base that can easily be used to
mount an FPGA board using an additional flat rectangular piece of Plexiglas or
Lexan. The rectangular sheet of plastic bolts to the base using the four screw
holes above the open cargo area, and the FPGA board bolts to the sheet of
plastic. Power can be provided at up to 1 amp using the iCreate’s internal
battery or there is also space in the cargo area for an additional battery pack for
the FPGA board. The internal battery voltage is 18V, so a 7, 9, or 5 volt DC
regulator (depends on which FPGA board) is needed to drop the voltage for the
FPGA board (note: too high of a DC input voltage will overheat the FPGA
board’s internal DC regulator). A DB-25 connector seen at the center in Figure
13.28 provides all of the power and serial port connections needed.

FPGA Robotics Projects 275

Figure 13.28 iRobots iCreate robot base can be controlled by an FPGA with a Nios Processor using
an RS-232 serial port to send commands to the iCreate robot’s internal microcontroller. Photograph
courtesy of iRobot.

Once you develop a working robot and want to run existing demos, you may
want to program the FPGA board’s flash memory configuration device so that
your design automatically runs whenever the board is turned on.

13.9 For Additional Information

Radio-controlled cars and parts such as batteries, battery chargers, and servos can be
obtained at a local hobby shop or via mail order at a lower cost from:

Tower Hobbies http://www.towerhobbies.com
P.O. Box 9078
Champaign, IL 61826-9078
800-637-6050

IR Proximity, Line Tracker, Sonar sensors, Servos, Wheels, and Robot kits can be
obtained via mail order from:

Lynxmotion, Inc. http://www.lynxmotion.com

104 Partridge Road

Pekin, IL 61554-1403

309-382-1254

276 Rapid Prototyping of Digital Systems Chapter 13

Mondotronics http://www.robotstore.com
4286 Redwood Highway #226
San Raphael, CA 94903
800–374–5764

Sensors & Robot kits, Servo wheels for robots, and Servo wheel encoder kits can also
be obtained via mail order from:

Acroname http://www.acroname.com
P.O. Box 1894
Nederland, CO 80466

303-258-3161

Low-cost digital and analog compass sensors are available via mail order from:

Dinsmore Instrument Co. http://www.dinsmoresensors.com

P.O. Box 345
Flint, Michigan 48501

810-744-1790

Electronic Compass Modules are available from:

PNI Corp http://www.pnicorp.com

5464 Skylane Blvd. Suite A

Santa Rosa, CA 95403

GPS and DGPS ICs and modules are available from:

Motorola TCG http://www.motorola.com/gps

GPS Products

2900 South Diablo Way

Tempe, AZ 85282

u-blox AG http://www.u-blox.com

Zürcherstrasse 68
8800 Thalwil
Schweiz

A wide array of robot sensor modules is available from:

Devantech Ltd (Robot Electronics) http://www.robot-electronics.co.uk
Unit 2B Gilray Road
Diss
Norfolk
IP22 4EU
England

FPGA Robotics Projects 277

Robotics Connection http://www.roboticsconnection.com
4355 Cobb Parkway
Suite J148
Atlanta, GA 30339

A wide array of various sensor and GPS boards is available from:

Spark Fun Electronics http://www.sparkfun.com
2500 Central Ave.
Suite Q
Boulder, CO 80301

A longer list of robot parts vendors and sites can be found at:

http://users.ece.gatech.edu/~hamblen/4006/robot_links.htm

13.10 Laboratory Exercises

1. Develop a counter design to find the dead zone of a converted R/C servo motor. The dead
or null zone is the time near 1.5ms that actually makes the servo motor stop moving. As
in the example motor driver code, send a width adjusted pulse every 20ms. You will need
a resolution of at least .01ms to find the dead zone, so a clock faster than the example
code is required. For example, the motor might actually stop at 1.54ms instead of 1.50ms.
Use the clk_div FPGAcore function to provide the clock. The design should increase the
width of the timing pulse if one pushbutton is hit and decrease the width if the other
pushbutton is hit. Display the width of the timing pulse in the seven-segment LEDs. Use
a Cyclone DIP-switch input to select the motor to examine. By hitting the pushbuttons,
you should be able to stop and reverse the motor. The dead zone will be between the
settings where the drive wheel reverses direction. At the dead zone, the drive wheel
should stop. Settings near the dead zone will make the motor run slower. Record the dead
zone for both the left and right motor.

2. Using the dead zone settings from problem 1, design a motor speed controller. Settings
within around .2ms of the dead zone will make the motor run slower. The closer to the
dead zone the slower the motor will run. Include at least four speed settings for each
motor. See if you can get the robot to move in a straight line at a slow speed.

3. Develop a speed controller for the robot drive motors by pulsing the drive motors on and
off. The motors are sent a pulse of 1ms for reverse and 2ms for forward at full speed. If
no pulse is sent for 20ms, the motor stops. If a motor is sent a 1 or 2ms pulse followed by
no pulse in a repeating pattern, it will move slower. To move even slower use pulse, no
pulse, no pulse in a repeating pattern. To move faster use pulse, no pulse, pulse in a
repeating pattern. Using this approach, develop a speed controller for the robot with at
least five speeds and direction. Send no pulse for the stop speed. Some additional
mechanical noise will result from pulsing the motors at slow speeds. See if the robot will
move in a straight line at a slow speed.

4. Use an IR LED and IR sensor to add position feedback to the motors. You can build it
yourself or a similar servo wheel encoder kit built by Nubotics is available from
Acroname. Some sensor modules are available that have both the IR LED and IR sensor
mounted in a single plastic case. For reflective sensors, mark the wheels with radial black

278 Rapid Prototyping of Digital Systems Chapter 13

paint stripes or black drafting tape and count the pulses from the IR sensor to determine
movement of the wheel. Another option would be to draw the radial stripes using a PC
drawing program and print it on clear adhesive labels made for laser printers. The labels
could then be placed on the flat side of the wheel. If a transmissive sensor arrangement is
used, holes can be drilled in the main wheel or a second smaller slotted wheel could be
attached to the servo output shaft that periodically interrupts the IR light beam from the
LED to the sensor. In this case, the LED and sensor are mounted on opposite sides of the
wheel. This same optical sensing technique is used in many mice to detect movement of
the mouse ball. Use the position feedback to implement more accurate variable speed and
position control for the motors.

5. Design a state machine using a counter/timer that will move the robot in the following
fixed pattern:

• Move forward for 6 seconds.

• Turn right and go forward for 4 seconds (do not count the time it takes to turn).

• Turn left and go forward for 2 seconds.

• Stop, pause for 2 seconds, turn 180 degrees, and start over.

Determine the amount of time required for 90- and 180-degree turns by trial and error. A
10Hz or 100Hz clock should be used for the timer. Use the clk_div FPGAcore to divide
the UP3 on-board clock. The state machine should check the timer to see if the correct
amount of time has elapsed before moving to the next state in the path. The timer is reset
when moving to a new portion of the path. Use an initial state that turns off the motors
until a pushbutton is hit, so that it is easier to control the robot during download. Since
there is no motor position feedback, all turns and the actual distance traveled by the
FPGA-bot will vary slightly.

Start 6 sec.
4 sec.
2 sec.
4 sec.

2 sec. Turn 180
Degrees
6 sec.

Figure 13.29 Simple path for state machine without sensor response.

6. Using a ROM, develop a ROM-based state machine that reads a motor direction and
time from the ROM. Put a complex pattern such as a dance step in the ROM using a
MIF file. For looping, another field in the ROM can be used to specify a jump to a
different next address.

FPGA Robotics Projects 279

7. Using the keyboard FPGAcore, design an interface to the keyboard that allows the
keyboard to be used as a remote control device to move the robot. Pick at least five
different keys to command to robot to move, turn left, turn right, or stop.

8. Interface an IR proximity sensor module to the FPGA-bot using jumpers connected
to the Cyclone male header socket. Attach the module in front of the header socket
using foam tape. Alternate driving the left and right IR LEDs at 100Hz. Check for an
IR sensor return and develop two signals, LEFT and RIGHT to indicate if the IR
sensor return is from the left or right IR LED. The IR LEDs may need to be adjusted
or shielded with some heat shrink tubing so that the floor does not reflect IR to the
sensor. Use the LEFT and RIGHT signals to drive the decimal points on two LEDs to
help adjust the sensor. It may be necessary to filter the IR returns using a counter
with a return/no return threshold for reliable operation. Using a clock faster than
100Hz, for example 10kHz, only set LEFT or RIGHT if the return was present for
several clock cycles.

9. Using IR sensor input, develop a design for the FPGA-bot that follows a person. The
person must be within a foot or so of the FPGA-bot. When a left signal is present
turn left, when a right signal is present turn right, and when both signals are present,
move forward a few inches and stop. When all signals are lost, the FPGA-bot should
rotate until an IR return is acquired.

10. Use motor speed control and a state machine with a timer to perform a small figure
eight with the FPGA-bot.

11. Once the IR proximity sensor module from problem 8 is interfaced, design a state
machine for the robot that moves forward and avoids obstacles. If it sees an obstacle
to the left, turn right, and if there is an obstacle to the right, turn left. If both left and
right obstacles are present, the robot should go backwards by reversing both motors.

12. With two FPGA-bots facing each other, develop a serial communications protocol
using the IR LEDs and sensors. Assume the serial data is fixed in length and always
starts with a known pattern at a fixed clock rate. The IR LEDs are pulsed at around
40kHz and the sensor has a 40kHz filter, so this will limit the bandwidth to a few
kHz. Transmit the 8-bit value from the Cyclone DIP switches and display the value
in the receiving FPGA-bot seven-segment LED displays. Display the raw IR sensor
input in the decimal point LED to aid in debugging and alignment.

13. Interface the line-following module to the FPGA-bot, and design a state machine that
follows a line. The line-following module has three sensor signals, left, center, and
right. If the line drifts to the left, turn right, and if the line drifts to the right, turn left.
Adjust turn constants so that the FPGA-bot moves along the line as fast as possible.
If speed control was developed for the FPGA-bot as suggested in earlier problems,
try using speed control for smaller less abrupt turns.

14. Using a standard IR remote control unit from a television or VCR and an IR sensor
interfaced to the FPGA-bot, implement a remote control for the FPGA-bot. Different
buttons on the remote control unit generate a different sequence of timing pulses. A
digital oscilloscope or logic analyzer can be used to examine the timing pulses.

15. Interface the a magnetic or electronic compass module to the FPGA-bot, and design a
state machine that performs the following operation:

• Turn North.

• Move forward 4 seconds.

280 Rapid Prototyping of Digital Systems Chapter 13

• Turn East.

• Move forward 4 seconds.

• Turn Southwest.

• Move forward 6.6 seconds.

• Stop and repeat when the pushbutton is hit.

The mechanical compass has a small time delay due to the inertia of the magnetized
rotor. Just like a real compass, it will swing back and forth for awhile before
stopping. With care, the leads on the compass module can be plugged into a DIP
socket with wire wrapped power supply and pull-up resistor connections on a small
protoboard or make a printed circuit board for the compass with jumper wires to plug
into the Cyclone female header socket. Make sure the compass module is mounted so
that it is level and as far away from the motors magnets as possible.

16. Interface a Sonar-ranging module to the FPGA-bot and perform the following
operation:

• Scan the immediate area 360 degrees by rotating the robot

• Locate the nearest object.

• Move close to the object and stop.

17. Attach the Sonar transducer to an unmodified servo’s output shaft. Use the new servo
to scan the area and locate the closest object. To sweep the unmodified servo back
and forth, a timing pulse that slowly increases from 1ms to 2ms and back to 1ms is
required. Move close to the nearest object and stop.

18. Attach several IR ranging sensors to the FPGA-bot and use the sensor data to
develop a wall following robot.

19. Interface additional sensors, switches, etc., to the FPGA-bot so that it can navigate a
maze. If several robots are being developed, consider a contest such as best time
through the maze or best time after learning the maze.

20. Use the μP 3 computer from Chapter 8 to implement a microcontroller to control the
robot instead of a custom state machine. Write a μP 3 assembly language program to
solve one of the previous problems. Interface a time-delay timer, the sensors, and the
motor speed control unit to the μP 3 computer using I/O ports as suggested in
problem 8.6. The additional machine instructions suggested in the exercises in
Chapter 8 would also be useful.

21. Use a Nios processor to control the robot with C code using the UP3 Nios II
reference design in Chapters 16 & 17.

FPGA Robotics Projects 281

22. Develop and hold a FPGA-bot design contest. Information on previous and current
robotics contests can be found online at various web sites. Here are some ideas that
have been used for other robot design contests:
• Robot Maze Solving

• Robot Dance Contest

• Sumo Wrestling

• Robot Soccer Teams

• Robot Laser Tag

• Fire Fighting Robots

• Robots that collect objects

• Robots that detect mines



CHAPTER 14

A RISC Design:
Synthesis of the MIPS
Processor Core

A full die photograph of the MIPS R2000 RISC Microprocessor is shown above. The
1986 MIPS R2000 with five pipeline stages and 450,000 transistors was the world’s first
commercial RISC microprocessor. Photograph ©1995-2004 courtesy of Michael
Davidson, Florida State University, http://micro.magnet.fsu.edu/chipshots.